Too important to fail.

OpenAI and Anthropic are spending staggering amounts of money, creating enormous debt among other things. The revenues are real, but the path to durable profits remains unclear. If you evaluate these companies as conventional businesses, you should be worried.

But a friend of mine, Ryan Finley, asked “What if these companies are already too important to fail?” Not too valuable. Not too promising. Too important.

Governments are using their products, defense agencies, large corporations, small organizations are using them to perform work they could never previously could. These institutions aren’t merely buying software, they are offloading parts of their intelligence and their workflows to a small number of AI companies.

Research. Analysis. Coding. Customer support. Logistics. Institutional knowledge. Decision-making. Thinking. Writing. Leadership. Lord knows what else our thoughtfully cheap society already offloaded to the robot.

Once enough critical work passes through these systems, the question changes. It is no longer simply whether the company has a healthy debt-to-revenue ratio. It becomes: What happens to everyone else if the company disappears?

Would the United States government allow one of its most strategically significant AI companies to dissolve? Would defense agencies permit a critical model provider to simply switch off? Would major corporations tolerate the sudden disappearance of systems woven into their workflows?

Nope.

We are assessing these companies with the wrong model. We look at their spending and ask whether they can become “normal sustainable businesses.” But perhaps “normal business” is no longer the relevant category. Perhaps frontier AI companies are becoming something closer to strategic infrastructure. Privately owned, enormously expensive, and now indispensable.

It’s already priced in, baby!